The Impact of Socioeconomic and Environmental Indicators on Economic Development: An Interdisciplinary Empirical Study
Round 1
Reviewer 1 Report
Overall, this is a technically impressive paper that seeks to apply a novel Structural Bayseian Panel Compressed Vector Autoregression model (SBPCVAR) to evaluate the relationship between environmental, health-related, and macroeconomic variables of interest in a panel of countries. The model is outlined reasonably well and theoretically can be useful, integrating the concepts of Bayesian VAR, Markov Chain Monte Carlo, and regressor shrinkage. However, I have got several significant concerns and potential improvement suggestions, mostly dealing with data-mining concerns, result presentation, and interpretations. They are listed below from most to least important.
1) The specification of the model leaves a lot of room for data-mining, and this study proposes little to no potential resolution for this. First, there is the probability threshold tau (p. 6) that is chosen to be 0.5%, which does not align with commonly chosen confidence intervals, and it is more restrictive than in prior literature which this study explicitly states. Second, the log natural Bayes factor (IBF) is another threshold for model selection that can be manipulated. This study uses IBF > 10 as strong evidence for submodel selection (and admits it is less restrictive than prior literature, p. 7). How one can select these thresholds? A more detailed discussion of why these methodological choices (using a less restrictive IBF criterion but a more restrictive PIP criterion) would be welcome. Alternatively, what the optimal model looks like if other values of tau or other IBF thresholds are considered? This would allow readers to assess the robustness of the results and potential applicability of the model more explicitly.
2) The interpretation and discussion of the results are quite brief, descriptive, and fragmented. First, little to no explicit discussion is devoted to the CPS (which gives the direction of the predictor effect on the outcome). For instance, why inflation and especially individuals using internet contributes to productivity negatively (predictors are significant based on PIP and have CPS < 0.5)? Second, the narrative on pp. 9-10 looks very generic and does not provide any further insights based on the results.
3) In the introduction, this study claims to use the model to interpret the implications of the GFC, the 2011-2014 recovery, the pandemic, and provide some insights on the macroeconomic implications of the currently ongoing Russia-Ukraine war (p. 3). However, little to no analysis or interpretation like this is present further in the paper.
4) There seem to be quite substantial multicollinearity concerns as the model seeks to predict productivity using weighted income per capita, GDP per capita PPP, and various other macroeconomic indicators that are quite tightly correlated.
5) Table 1 formatting could be improved. First, "poorly-paid job" and "employment to population ratio" predictors have got inputs in wrong columns. Second, the notes mention that "the last row refers to the outcomes of interest at time t corresponding to the economic growth". However, the last row just outlines the variable name (productivity) and abbreviation (lgdp) without any further elaboration which is confusing. Third, the same notes state GDP growth is computed using a logarithm of a ratio between GDP per capita and population. This might be a typo, but it adds to the overall confusion when trying to comprehend the results.
Author Response
Please, see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Thank you for the opportunity to review the article, the reasoning of which was presented in a clear way. I am also pleased with the interdisciplinarity of the undertaken research.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The paper investigates an interesting topic about the effects of environment source and health statistics on economic growth. The SPBVAR model is applied to select the variables and the causal relationship is obtained between sustainability indicators and economic growth.
Here are a few major concerns before the decision can be made:
1. The author has to show more statistical results about the model, like parameter estimation and statistical testing.
2. More description about PIP should be added since it is an important statistic to be used in the paper to select variables, like how to decide the cutoff value tau.
3. Elaborate more details about how the casual inference is achieved in this paper.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
Some practical examples should be displayed better in research.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Overall, the revised version of the manuscript reads much better and conveys the applicability of the model and the interpretation of its results more clearly. I would like to congratulate the author(s) on their hard work during the revision process and for accommodating the suggestions. However, I have got several further comments and suggestions that I believe still need addressing, listed below from most to least important.
1) After the specification of the model was changed slightly (PIP increased to 1% and IBF reduced to 8), the CPS for internet use (variable 24) is showing a strong positive relationship (contrastingly with strong negative relationship in a specification with PIP of 0.5% and IBF of 10, prior version of the manuscript). This is quite alarming for the robustness of model application, therefore I would suggest expanding on the stability of model results and predictions.
2) Table 3 reports a Sargan test statistic significant at 1%. This can generally be interpreted as the rejection of the null hypothesis of instrument validity. Nevertheless, the paper claims this is a sign of result robustness. A further clarification on the Sargan test interpretation would then be welcome.
3) Table 2 seems to have several formatting errors. First, the inflation coefficient equals -0.96 and is reported to be significant at 5% with a standard error of 0.86. Either the standard error is incorrect, or statistical significance is misreported. Second, the coefficients for variables 22 and 23 seem to be duplicated. Please consider revising the formatting of Table 2.
4) Productivity is still being measured as the log of GDP PPP per capita divided by population. This is not very meaningful, as the GDP PPP is effectively being divided by population twice. It is either a typo or a calculation error (comment carrying forward from prior revision).
5) Although the PIP and IBF thresholds have been changed an it is correctly reflected in Section 2, there are legacy mentions of PIP and IBF being more or less restrictive than prior literature (p. 6). Please consider proofreading.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Thanks for taking the effort to address my concerns.
Author Response
Please, see the attachment.
Author Response File: Author Response.pdf
